Neuromorphic computing that mimics the human brain (opens in new tab) is one step closer to reality, as Washington State University researchers have built a crucial circuit for this new type of computing using an implausibly pure substance.
Using bee honey, the researchers built a proof-of-concept memory resistor, or memristor. To accomplish this feat, they first turned the honey into a solid form and then held it between two metal electrodes in the same way the brain’s synapses lie between pairs of neurons.
After its inception, researchers at Washington State University tested the device’s ability to turn on and off quickly at speeds between 100 and 500 nanoseconds. The tests were successful, and the researchers hope their new memristor can help pave the way for biodegradable, sustainable and organically based computing systems in the future.
In a press release (opens in new tab) Announcing the discovery, Feng Zhao, associate professor of WSU’s School of Engineering and Computer Science, gave more insight into honey’s potential in making brain-like computer chips, saying:
“This is a very small device with a simple structure, but it has very similar functionalities to a human neuron. This means that if we can integrate millions or billions of these honey memristors together, they can be turned into a neuromorphic system that is much like a human brain.”
Conventional computer systems such as those found in business computers (opens in new tab) and mobile workstations (opens in new tab) are based on the von Neumann architecture that includes an input such as a keyboard and mouse, as well as an output such as a monitor, along with a CPU and RAM.
Link: The beginning of this link looks broken. The mechanisms from input to processing to memory to output takes much more power compared to the human brain. For example Fugaku . from Fujitsu (opens in new tab) supercomputer uses 28 million watts to run, while the human brain only uses about 10 to 20 watts. This is why companies like Intel and IBM are working on neuromorphic chips (opens in new tab) that mimic how the human brain functions.
The human brain has over 100 billion neurons with over 1,000 tons of synapses or connections between them. Because each neuron can both process and store data, the brain is much more efficient than a traditional computer.
At the same time, conventional computer chips (opens in new tab) are built using non-renewable and toxic materials, while neuromorphic chips, such as those made by Washington State University researchers, can instead be made with biodegradable materials.
In the future, Zhao’s team wants to reduce the size of its honey memristors from a microscale that is about the size of a human hair to a nanoscale that is about 1/1000 of a human hair. By doing so, the researchers will be able to bundle together millions or even billions of honey memristors to create an entire neuromorphic computer system.